Re: [RFC PATCH] ceph: initialize superblock s_time_gran to 1

[Luis Henriques <lhenriques@suse.com>]

Jeff Layton <jlayton@kernel.org> writes:

> On Thu, 2019-06-27 at 15:44 +0000, Sage Weil wrote:
>> On Thu, 27 Jun 2019, Jeff Layton wrote:
>> > On Thu, 2019-06-27 at 14:51 +0100, Luis Henriques wrote:
>> > > Having granularity set to 1us results in having inode timestamps with a
>> > > accurancy different from the fuse client (i.e. atime, ctime and mtime will
>> > > always end with '000').  This patch normalizes this behaviour and sets the
>> > > granularity to 1.
>> > > 
>> > > Signed-off-by: Luis Henriques <lhenriques@suse.com>
>> > > ---
>> > >  fs/ceph/super.c | 2 +-
>> > >  1 file changed, 1 insertion(+), 1 deletion(-)
>> > > 
>> > > Hi!
>> > > 
>> > > As far as I could see there are no other side-effects of changing
>> > > s_time_gran but I'm really not sure why it was initially set to 1000 in
>> > > the first place so I may be missing something.
>> > > 
>> > > diff --git a/fs/ceph/super.c b/fs/ceph/super.c
>> > > index d57fa60dcd43..35dd75bc9cd0 100644
>> > > --- a/fs/ceph/super.c
>> > > +++ b/fs/ceph/super.c
>> > > @@ -980,7 +980,7 @@ static int ceph_set_super(struct super_block *s, void *data)
>> > >     s->s_d_op = &ceph_dentry_ops;
>> > >     s->s_export_op = &ceph_export_ops;
>> > >  
>> > > -   s->s_time_gran = 1000;  /* 1000 ns == 1 us */
>> > > +   s->s_time_gran = 1;
>> > >  
>> > >     ret = set_anon_super(s, NULL);  /* what is that second arg for? */
>> > >     if (ret != 0)
>> > 
>> > 
>> > Looks like it was set that way since the client code was originally
>> > merged. Was this an earlier limitation of ceph that is no longer
>> > applicable?
>> > 
>> > In any case, I see no need at all to keep this at 1000, so:
>> 
>> As long as the encoded on-write time value is at ns resolution, I 
>> agree!  No recollection of why I did this :(
>> 
>> Reviewed-by: Sage Weil <sage@redhat.com>
>
> Good enough for me. I went ahead and merged this into the testing
> branch. Assuming nothing breaks, this should make v5.3.

Awesome, thanks.  AFAICS it shouldn't break anything, specially because
the fuse client seems to be using ns resolution too.  But yeah
unexpected side-effects show up in unexpected ways :-)

Cheers,
-- 
Luis